Internalized Bloch matrix

[pablosanjose]

This change removes all the similarmatrix API. Having to allocate a matrix before being able to have a reusable bloch! is cumbersome. This PR partially restores the older design changed in pablosanjose/Elsa.jl#28

The similarmatrix is allocated upon creating a hamiltonian and its type can now be controlled with the keyword blochtype. We can thus have a multiorbital h::Hamiltonian that carries inside a flat blochmatrix to be returned when calling bloch!(h, ...). To avoid allocating such a matrix at each step of any transformation, we carry over the same matrix at any transformation step that preserves its size and eltype.

We can then now do

h = LatticePresets.honeycomb() |> hamiltonian(hopping(I), orbitals = (Val(2), Val(1)), blochtype = Matrix{ComplexF64})
bloch!(h, (1,2)) # returns a flattened 3x3 Matrix{ComplexF64}

We have also the special syntax

h = LatticePresets.honeycomb() |> hamiltonian(hopping(I), orbitals = (Val(2), Val(1)), blochtype = flatten)
bloch!(h, (1,2)) # returns a flattened 3x3 SparseMatrixCSC{ComplexF64}

Together with this we can also change the blochtype of a Hamiltonian or ParametricHamiltonian h with

hamiltonian(h; blochtype = newtype)

Another benefit is that we get rid of the pesky optimize! function. The (Parametric)Hamiltonian harmonics don't need to pre-store any structural zeros to avoid sparse splicing. We just optimize the internalized blochmatrix upon creation, in accordance with the elements of all harmonics (function _similar_merged_sparse now).

This opens the door to having h(ϕs; params...) as a shorthand for bloch(h(; params...), ϕs), or even bloch!(h(; params...), ϕs). This is not implemented here yet, but could be used to even stop exporting bloch and bloch!. Even further, h(not_all_phis; params...) could even replace wrap, but that's another issue.

[codecov-commenter]

Codecov Report

Merging #105 into master will decrease coverage by 0.04%.
The diff coverage is 86.11%.

@@            Coverage Diff             @@
##           master     #105      +/-   ##
==========================================
- Coverage   62.37%   62.32%   -0.05%     
==========================================
  Files          16       16              
  Lines        2708     2734      +26     
==========================================
+ Hits         1689     1704      +15     
- Misses       1019     1030      +11     

Impacted Files	Coverage Δ
src/Quantica.jl	100.00% <ø> (ø)
src/greens.jl	0.00% <ø> (ø)
src/model.jl	72.66% <ø> (ø)
src/plot_vegalite.jl	0.00% <0.00%> (ø)
src/tools.jl	56.54% <75.00%> (-0.91%)	⬇️
src/diagonalizer.jl	52.71% <81.81%> (-1.43%)	⬇️
src/hamiltonian.jl	74.76% <84.55%> (-0.20%)	⬇️
src/KPM.jl	64.80% <100.00%> (ø)
src/bandstructure.jl	91.15% <100.00%> (-0.06%)	⬇️
src/lattice.jl	81.59% <100.00%> (ø)
... and 4 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 743b928...28e52f1. Read the comment docs.